Odd behaviour after system outage

[Randy_B]

I have a question about some odd behaviour we had over the weekend on one our our ESX 3.5 clusters. We had a power related outage (as always, in the middle of the night) and all the hosts in a cluster of 5 BL680's went down. The cluster was set for fully automated. When the first two hosts were brought up, all the VM's in the cluster (180) powered up automatically on those first two hosts. These are 16 core servers and 90 VM's actually ran on each box but the service console and any tasks in VC were hosed until we powered down about 20 of the guests as the CPU for the console was pegged. Trying to Vmotion or shut down machines would just hang there.

What I don't understand is why did the guests automatically start up? I didn't think the fully automated setting would do this and none of my hosts are set to automatically start up the guest with the system.

Thanks,

Randy

[letoatrads]

Sounds like the default behaviour of VMWare HA - do you have HA enabled on your cluster? ( Right-click on cluster / Edit Settings -first page )

[Randy_B]

I do have HA and DRS enabled on the cluster but not just one or two hosts went down, the whole cluster went down. When the first two hosts were powered up, all 180 VMs powered up on those two hosts. I've been through datacenter blackouts before with ESX 3.x and never seen that before.

[Troy_Clavell]

quick question... What is your admission control set to on HA?

[letoatrads · Accepted solution]

On the HA Cluster options -

Do not power on VM's if they violate availitibilty constraints

vs

Allow VM's to be powered on if they violate constraints

If you have the second option chosen, HA will bring up ALL VM's as soon as a single host is available - regardless of whether the Host can handle the load.

[Randy_B]

That was it, the cluster was set to allow violations. Thanks.